You are on page 1of 1

10.

6 Discussion and Outlook 583

If one insists on interpreting the behaviour of the Hubbard model in


terms of an effective Hamiltonian, one has to study a t-J model with
“parametrically small” values34of J/t [328].
A t-J model with correctly scaled coefficients is (dropping the den-
sity-density interaction)

-.
(10.75)
where we used the Hubbard operators (5.21) to express the kinetic en-
ergy. The projected hopping term still keeps the on-site dynamics, and
the problem remains difficult even at J = 0 where we recover the U = 00
Hubbard model. It is, however, a welcome simplification that the limit
D -+ 00 allows to replace the exchange term with its static Hartree-
decoupled form

-C Si
J
* Sj = C JqSq * S-q C JqSq * (S-q) (10.76)
2D (iJ> 9 q

where the Fourier component Jq is

Jq =
J D
-
D j=1
ccos ( q p ) (10.77)

It follows that in its paramagnetic phase, the t-J model becomes iden-
tical to the U = 00 Hubbard model.
Noticing that the decoupling (10.76) is analogous to the decoupling
of the Hubbard term in (7.37), and that a mean field treatment is now
strictly justifiable, we can make a derivation parallel to that in Sec. 7.5
to arrive at the q-dependent susceptibility of the t-J model [328]

(10.78)

This expression has the same structure as the RPA result (7.44). Notice,
however, that the role of the free-electron susceptibility has been taken
341t is a different question that one can postulate a large-D t-J model with a
different parameter range [59].

You might also like